Mikey Does

Why Collocations Are the Hardest Part of Japanese Vocabulary (And What to Do About It)

Most Japanese learners hit a wall somewhere between N4 and N2. They have a large vocabulary. They understand grammar. They can read sentences and follow conversations. But when they generate Japanese — write an email, speak to a native speaker, produce their own sentences — something is off. The words are right but the combinations aren’t quite natural. Native speakers understand them, but there’s friction. Sometimes they correct you. More often they don’t bother.

The culprit is almost always the same: collocational knowledge. Not vocabulary size. Not grammar errors. The specific, semi-predictable ways that Japanese words prefer to co-occur with other words — preferences that are real, consistent, and learnable, but that most learning approaches address only by accident.

What Is a Collocation?

A collocation is a word pair or phrase whose combination occurs at higher-than-chance frequency. The individual words are familiar; the preferred pairing is what you have to learn separately.

In English: you “make a decision” — not “do a decision” or “take a decision.” You “break a record,” not “smash a record” (though “smash a record” is understandable). “Strong tea” is correct; “powerful tea” sounds weird.

These aren’t random. They’re statistically patterned combinations that native speakers have absorbed through massive input exposure and that non-native speakers commonly get wrong — because knowing the words individually doesn’t tell you how they prefer to combine.

In Japanese, collocations operate at several levels:

Verb collocations (verb + noun; adjective + verb):

  • 決断を下す (ketsudan wo kudasu) — “to make a decision” (literally “lower a decision”) — kudasu is the preferred verb, not suru in formal contexts
  • 問題が生じる (mondai ga shōjiru) — “a problem arises” — shōjiru is the natural verb for this; okiru would feel wrong to a native speaker
  • 結論に達する (ketsuron ni tassuru) — “to reach a conclusion” — tassuru is the expected verb

Adjective collocations:

  • 深い眠り (fukai nemuri) — “deep sleep” — fukai (deep) collocates with sleep; using ooi (a different “much/great”) would be wrong
  • 強い影響 (tsuyoi eikyō) — “strong influence” — tsuyoi is preferred; using ōkii influence (big influence) changes the register

Verb + particle collocations:

  • 意見を言う vs. 意見を述べる (iken wo iu vs. iken wo noberu) — you can “say” an opinion informally, but formal registers expect noberu (to state/express)
  • 電話をかける (denwa wo kakeru) — “to make a phone call” — kakeru is the fixed collocate for telephone; you can’t substitute another verb

Register-specific collocations (particularly important in Japanese):

Japanese’s deep formal/informal register system means that even when the semantic content is the same, the preferred word combinations shift. The verb you use with “meeting” in business writing (kaiigi ga kaisai sareru) differs from casual speech (kaiigi ga aru) — and using the casual collocation in a business email will be noticed.

Why Textbooks Miss This

Standard Japanese learning materials — textbooks, frequency lists, SRS decks based on individual words — are structured around individual lexical items, not lexical chunks. When you learn 決断 (ketsudan, “decision”), you’re presented with the word and its meaning. The textbook’s vocabulary section does not tell you that in formal writing, 決断を下す is the natural construction; you would have to deduce or encounter this yourself.

The challenge is that collocational patterns require corpus-level data to teach properly. A textbook author writing from intuition may include natural examples, but without systematic analysis of how words actually distribute in real text, the coverage is incomplete and idiosyncratic.

This is why intermediate learners often plateau: their individual vocabulary is growing but their collocational knowledge isn’t keeping pace. They know the words but not how those words prefer to travel.

The Research Case for Collocations

Applied linguistics researcher Michael Lewis coined the term “collocational knowledge” as a distinct component of vocabulary knowledge in his 1993 and 1997 books arguing for the Lexical Approach — the idea that language learning should focus on learning chunks and collocations, not individual words.

The research supports the importance of this gap:

  • Nesselhauf (2005) studied German university students’ English collocations in academic writing and found that verb-noun collocations were the most problematic error category — even for advanced learners.
  • Nation (2001) in his model of vocabulary knowledge separated “form” (the word) from “collocation” (how it combines with other words), treating them as distinct developmental stages. A learner can know a word at the level of form without knowing its collocational behavior.
  • Studies of L2 Japanese specifically (e.g., Kita 2011, Miyamoto 2016) have found that intermediate-to-advanced Japanese learners systematically underuse higher-register collocations and overuse the semantically equivalent lower-register forms — exactly the pattern produced by learning individual words without attention to combinatorial preferences.

How to Actually Build Collocational Knowledge

1. Read sentences, not words

The single most important thing is consistently learning vocabulary in sentence context — specifically, sentences from native materials where the collocation is visible. A flashcard with ケッタン on one side and “decision” on the other teaches you a word. A flashcard with 重大な決断を下した teaches you a real-world chunk.

This is why sentence mining — creating SRS cards from sentences you encounter in native text — builds collocational knowledge where word-list drilling doesn’t. The sentence preserves the collocational environment.

If you use Sakubo for SRS, importing sentence cards from native sources (manga, articles, novels) rather than dictionary entries means each review is also reinforcing the collocation.

2. Use a Japanese collocations dictionary

The 明鏡国語辞典 (Meikyo) and 新明解国語辞典 (Shinmeikai) both include more contextual usage information than learner-focused dictionaries. For collocations specifically:

  • JMdict / Jisho.org includes some collocational notes, but coverage is inconsistent.
  • The BCCWJ (Balanced Corpus of Contemporary Written Japanese) — available via the NINJAL online search interface — lets you search for collocation patterns in 100 million words of real Japanese text. Slightly tool-heavy but invaluable for checking “does this verb go with this noun?”

3. Notice the pattern when you encounter it

Active noticing — consciously flagging “this verb is strange/interesting next to this noun” when you encounter a collocation in native text — accelerates acquisition significantly compared to passive exposure. Linguist Dick Schmidt called this noticing: the threshold between input and intake. You see 意見を述べる, you know 言う is the obvious verb choice, and you consciously note that noberu is the formal preference. That moment of noticing dramatically increases the chance it sticks.

4. Mine dense collocation sources

Some text types are particularly collocation-dense relative to their vocabulary difficulty:

  • Business Japanese (敬語, keigo, and formal writing) — densely formulaic; collocations are extremely consistent and high-value to learn
  • News Japanese (NHK Web Easy is good for lower-levels; regular NHK for advanced) — constrained register, consistent verb-noun pairings
  • JLPT N2/N1 prep materials — although not a literary source, JLPT reading passages are specifically chosen to include collocations that distinguish near-native from merely proficient

The Pattern to Watch For

A quick diagnostic: if you’re an intermediate Japanese learner, try writing a short formal email in Japanese (requesting a meeting, thanking someone for a favor). Then compare your draft against a Japanese template from a native source. The differences — where you used suru and the native used nasaru or itasu, where you wrote ありがとうございます and the template has お礼申し上げます, where your verb choice was understood but slightly off-register — are almost entirely collocational. Not grammar. Not vocabulary gaps. Combinations.

That gap is your collocational knowledge gap. It’s fixable, but only through exposure to real native input at scale and deliberate attention to how words travel together.

Related Reading

Sources